[117592] in Cypherpunks
RE: Build a better OTP?
daemon@ATHENA.MIT.EDU (Lucky Green)
Mon Sep 6 19:28:10 1999
From: "Lucky Green" <shamrock@cypherpunks.to>
To: <cypherpunks@cyberpass.net>
Date: Mon, 6 Sep 1999 16:17:16 -0700
Message-ID: <NDBBIFGOKODBCKDGJDKLEEKPCFAA.shamrock@cypherpunks.to>
MIME-Version: 1.0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
In-Reply-To: <19990906212016.30353.qmail@nym.alias.net>
Reply-To: "Lucky Green" <shamrock@cypherpunks.to>
What is it about this recent tendency on Cypherpunks to cite "Authorities",
rather than bothering to think things through for more than a few seconds?
Weird...
Anyway, what Ben and Paul analyzed were the design assumptions, the design,
and data provided to them by Intel. So let's assume Ben and Paul's findings
are correct, which by the way I do. Do these findings imply that the output
of Intel's RNG is in fact random? Of course not.
In an ideal world, theory and practice are identical. Engineers - and
certainly cryptographers - don't operate in such a world. Chip manufacturers
do neither: chip designers make mistakes. See the listings of what Intel
calls "errata". Even if the chip design process was perfect, which it is
not, chip manufacturing is far from perfect. Any implementation errors
aside, some of the millions of CPU's shipping with RNG's will have
manufacturing defects. Some of these defects will affect the RNG. Some of
the defects affecting the RNG will make it fail invisibly, unnoticed by the
user absent access to the pre-whitening output. Which Intel refuses to
provide.
What makes Intel's bizarre refusal to provide access to the pre-whitening
output of the RNG all the more puzzling is that there is absolutely no harm
in doing so. While much security would be gained. Intel's seemingly
irrational refusal simply makes no sense to this observer at this time.
--Lucky Green <shamrock@cypherpunks.to>
> -----Original Message-----
> From: owner-cypherpunks@Algebra.COM
> [mailto:owner-cypherpunks@Algebra.COM]On Behalf Of lcs Mixmaster
> Remailer
> Sent: Monday, September 06, 1999 14:20
> To: cypherpunks@cyberpass.net
> Subject: Re: Build a better OTP?
>
>
>
> > >> Has anyone designed a true random number generator that runs at high
> > >> speed? Something that could fill a CD or DVD in a reasonable
> amount of
> > >> time?
> >
> > >Yes, there's one out by a little company called Intel. You might have
> > >heard of them.
> >
> > Yes but is it trustworthy?
>
> Your naive paranoia would be charming if it weren't so tiresome.
>
> The basic design is sound; see http://www.cryptography.com/intelRNG.pdf
> for a review by Ben Jun and Paul Kocher, two of the smartest guys around
> when it comes to real-world implementations of crypto technology:
